Merge from FreeBSD 2003/07/15 15:49:53 PDT commit to sys/netinet.
[dragonfly.git] / sys / kern / kern_exec.c
CommitLineData
984263bc
MD
1/*
2 * Copyright (c) 1993, David Greenman
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 *
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24 * SUCH DAMAGE.
25 *
26 * $FreeBSD: src/sys/kern/kern_exec.c,v 1.107.2.15 2002/07/30 15:40:46 nectar Exp $
3b568787 27 * $DragonFly: src/sys/kern/kern_exec.c,v 1.6 2003/06/26 05:55:14 dillon Exp $
984263bc
MD
28 */
29
30#include <sys/param.h>
31#include <sys/systm.h>
32#include <sys/sysproto.h>
33#include <sys/kernel.h>
34#include <sys/mount.h>
35#include <sys/filedesc.h>
36#include <sys/fcntl.h>
37#include <sys/acct.h>
38#include <sys/exec.h>
39#include <sys/imgact.h>
40#include <sys/imgact_elf.h>
41#include <sys/wait.h>
42#include <sys/malloc.h>
43#include <sys/proc.h>
44#include <sys/signalvar.h>
45#include <sys/pioctl.h>
46#include <sys/namei.h>
47#include <sys/sysent.h>
48#include <sys/shm.h>
49#include <sys/sysctl.h>
50#include <sys/vnode.h>
51#include <sys/aio.h>
52
53#include <vm/vm.h>
54#include <vm/vm_param.h>
55#include <sys/lock.h>
56#include <vm/pmap.h>
57#include <vm/vm_page.h>
58#include <vm/vm_map.h>
59#include <vm/vm_kern.h>
60#include <vm/vm_extern.h>
61#include <vm/vm_object.h>
62#include <vm/vm_pager.h>
63
64#include <sys/user.h>
65#include <machine/reg.h>
66
67MALLOC_DEFINE(M_PARGS, "proc-args", "Process arguments");
68
69static register_t *exec_copyout_strings __P((struct image_params *));
70
71/* XXX This should be vm_size_t. */
72static u_long ps_strings = PS_STRINGS;
73SYSCTL_ULONG(_kern, KERN_PS_STRINGS, ps_strings, CTLFLAG_RD, &ps_strings, 0, "");
74
75/* XXX This should be vm_size_t. */
76static u_long usrstack = USRSTACK;
77SYSCTL_ULONG(_kern, KERN_USRSTACK, usrstack, CTLFLAG_RD, &usrstack, 0, "");
78
79u_long ps_arg_cache_limit = PAGE_SIZE / 16;
80SYSCTL_LONG(_kern, OID_AUTO, ps_arg_cache_limit, CTLFLAG_RW,
81 &ps_arg_cache_limit, 0, "");
82
83int ps_argsopen = 1;
84SYSCTL_INT(_kern, OID_AUTO, ps_argsopen, CTLFLAG_RW, &ps_argsopen, 0, "");
85
86/*
87 * Each of the items is a pointer to a `const struct execsw', hence the
88 * double pointer here.
89 */
90static const struct execsw **execsw;
91
92#ifndef _SYS_SYSPROTO_H_
93struct execve_args {
94 char *fname;
95 char **argv;
96 char **envv;
97};
98#endif
99
100/*
101 * execve() system call.
102 */
103int
41c20dac 104execve(struct execve_args *uap)
984263bc 105{
dadab5e9
MD
106 struct thread *td = curthread;
107 struct proc *p = td->td_proc;
984263bc
MD
108 struct nameidata nd, *ndp;
109 register_t *stack_base;
110 int error, len, i;
111 struct image_params image_params, *imgp;
112 struct vattr attr;
113 int (*img_first) __P((struct image_params *));
114
dadab5e9 115 KKASSERT(p);
984263bc
MD
116 imgp = &image_params;
117
118 /*
119 * Lock the process and set the P_INEXEC flag to indicate that
120 * it should be left alone until we're done here. This is
121 * necessary to avoid race conditions - e.g. in ptrace() -
122 * that might allow a local user to illicitly obtain elevated
123 * privileges.
124 */
125 p->p_flag |= P_INEXEC;
126
127 /*
128 * Initialize part of the common data
129 */
130 imgp->proc = p;
131 imgp->uap = uap;
132 imgp->attr = &attr;
133 imgp->argc = imgp->envc = 0;
134 imgp->argv0 = NULL;
135 imgp->entry_addr = 0;
136 imgp->vmspace_destroyed = 0;
137 imgp->interpreted = 0;
138 imgp->interpreter_name[0] = '\0';
139 imgp->auxargs = NULL;
140 imgp->vp = NULL;
141 imgp->firstpage = NULL;
142 imgp->ps_strings = 0;
143
144 /*
145 * Allocate temporary demand zeroed space for argument and
146 * environment strings
147 */
148 imgp->stringbase = (char *)kmem_alloc_wait(exec_map, ARG_MAX + PAGE_SIZE);
149 if (imgp->stringbase == NULL) {
150 error = ENOMEM;
151 goto exec_fail;
152 }
153 imgp->stringp = imgp->stringbase;
154 imgp->stringspace = ARG_MAX;
155 imgp->image_header = imgp->stringbase + ARG_MAX;
156
157 /*
158 * Translate the file name. namei() returns a vnode pointer
159 * in ni_vp amoung other things.
160 */
161 ndp = &nd;
162 NDINIT(ndp, LOOKUP, LOCKLEAF | FOLLOW | SAVENAME,
dadab5e9 163 UIO_USERSPACE, uap->fname, td);
984263bc
MD
164
165interpret:
166
167 error = namei(ndp);
168 if (error) {
169 kmem_free_wakeup(exec_map, (vm_offset_t)imgp->stringbase,
170 ARG_MAX + PAGE_SIZE);
171 goto exec_fail;
172 }
173
174 imgp->vp = ndp->ni_vp;
175 imgp->fname = uap->fname;
176
177 /*
178 * Check file permissions (also 'opens' file)
179 */
180 error = exec_check_permissions(imgp);
181 if (error) {
dadab5e9 182 VOP_UNLOCK(imgp->vp, 0, td);
984263bc
MD
183 goto exec_fail_dealloc;
184 }
185
186 error = exec_map_first_page(imgp);
dadab5e9 187 VOP_UNLOCK(imgp->vp, 0, td);
984263bc
MD
188 if (error)
189 goto exec_fail_dealloc;
190
191 /*
192 * If the current process has a special image activator it
193 * wants to try first, call it. For example, emulating shell
194 * scripts differently.
195 */
196 error = -1;
197 if ((img_first = imgp->proc->p_sysent->sv_imgact_try) != NULL)
198 error = img_first(imgp);
199
200 /*
201 * Loop through the list of image activators, calling each one.
202 * An activator returns -1 if there is no match, 0 on success,
203 * and an error otherwise.
204 */
205 for (i = 0; error == -1 && execsw[i]; ++i) {
206 if (execsw[i]->ex_imgact == NULL ||
207 execsw[i]->ex_imgact == img_first) {
208 continue;
209 }
210 error = (*execsw[i]->ex_imgact)(imgp);
211 }
212
213 if (error) {
214 if (error == -1)
215 error = ENOEXEC;
216 goto exec_fail_dealloc;
217 }
218
219 /*
220 * Special interpreter operation, cleanup and loop up to try to
221 * activate the interpreter.
222 */
223 if (imgp->interpreted) {
224 exec_unmap_first_page(imgp);
225 /* free name buffer and old vnode */
226 NDFREE(ndp, NDF_ONLY_PNBUF);
227 vrele(ndp->ni_vp);
228 /* set new name to that of the interpreter */
229 NDINIT(ndp, LOOKUP, LOCKLEAF | FOLLOW | SAVENAME,
dadab5e9 230 UIO_SYSSPACE, imgp->interpreter_name, td);
984263bc
MD
231 goto interpret;
232 }
233
234 /*
235 * Copy out strings (args and env) and initialize stack base
236 */
237 stack_base = exec_copyout_strings(imgp);
238 p->p_vmspace->vm_minsaddr = (char *)stack_base;
239
240 /*
241 * If custom stack fixup routine present for this process
242 * let it do the stack setup.
243 * Else stuff argument count as first item on stack
244 */
245 if (p->p_sysent->sv_fixup)
246 (*p->p_sysent->sv_fixup)(&stack_base, imgp);
247 else
248 suword(--stack_base, imgp->argc);
249
250 /*
251 * For security and other reasons, the file descriptor table cannot
252 * be shared after an exec.
253 */
254 if (p->p_fd->fd_refcnt > 1) {
255 struct filedesc *tmp;
256
257 tmp = fdcopy(p);
258 fdfree(p);
259 p->p_fd = tmp;
260 }
261
262 /*
263 * For security and other reasons, signal handlers cannot
264 * be shared after an exec. The new proces gets a copy of the old
265 * handlers. In execsigs(), the new process will have its signals
266 * reset.
267 */
268 if (p->p_procsig->ps_refcnt > 1) {
269 struct procsig *newprocsig;
270
271 MALLOC(newprocsig, struct procsig *, sizeof(struct procsig),
272 M_SUBPROC, M_WAITOK);
273 bcopy(p->p_procsig, newprocsig, sizeof(*newprocsig));
274 p->p_procsig->ps_refcnt--;
275 p->p_procsig = newprocsig;
276 p->p_procsig->ps_refcnt = 1;
277 if (p->p_sigacts == &p->p_addr->u_sigacts)
278 panic("shared procsig but private sigacts?");
279
280 p->p_addr->u_sigacts = *p->p_sigacts;
281 p->p_sigacts = &p->p_addr->u_sigacts;
282 }
283
284 /* Stop profiling */
285 stopprofclock(p);
286
287 /* close files on exec */
288 fdcloseexec(p);
289
290 /* reset caught signals */
291 execsigs(p);
292
293 /* name this process - nameiexec(p, ndp) */
294 len = min(ndp->ni_cnd.cn_namelen,MAXCOMLEN);
295 bcopy(ndp->ni_cnd.cn_nameptr, p->p_comm, len);
296 p->p_comm[len] = 0;
297
298 /*
299 * mark as execed, wakeup the process that vforked (if any) and tell
300 * it that it now has its own resources back
301 */
302 p->p_flag |= P_EXEC;
303 if (p->p_pptr && (p->p_flag & P_PPWAIT)) {
304 p->p_flag &= ~P_PPWAIT;
305 wakeup((caddr_t)p->p_pptr);
306 }
307
308 /*
309 * Implement image setuid/setgid.
310 *
311 * Don't honor setuid/setgid if the filesystem prohibits it or if
312 * the process is being traced.
313 */
314 if ((((attr.va_mode & VSUID) && p->p_ucred->cr_uid != attr.va_uid) ||
315 ((attr.va_mode & VSGID) && p->p_ucred->cr_gid != attr.va_gid)) &&
316 (imgp->vp->v_mount->mnt_flag & MNT_NOSUID) == 0 &&
317 (p->p_flag & P_TRACED) == 0) {
318 /*
319 * Turn off syscall tracing for set-id programs, except for
320 * root. Record any set-id flags first to make sure that
321 * we do not regain any tracing during a possible block.
322 */
41c20dac 323 setsugid();
dadab5e9 324 if (p->p_tracep && suser(td)) {
984263bc
MD
325 struct vnode *vtmp;
326
327 if ((vtmp = p->p_tracep) != NULL) {
328 p->p_tracep = NULL;
329 p->p_traceflag = 0;
330 vrele(vtmp);
331 }
332 }
333 /* Close any file descriptors 0..2 that reference procfs */
334 setugidsafety(p);
335 /* Make sure file descriptors 0..2 are in use. */
336 error = fdcheckstd(p);
337 if (error != 0)
338 goto exec_fail_dealloc;
339 /*
340 * Set the new credentials.
341 */
e9a372eb 342 cratom(&p->p_ucred);
984263bc 343 if (attr.va_mode & VSUID)
41c20dac 344 change_euid(attr.va_uid);
984263bc
MD
345 if (attr.va_mode & VSGID)
346 p->p_ucred->cr_gid = attr.va_gid;
347 } else {
41c20dac
MD
348 if (p->p_ucred->cr_uid == p->p_ucred->cr_ruid &&
349 p->p_ucred->cr_gid == p->p_ucred->cr_rgid)
984263bc
MD
350 p->p_flag &= ~P_SUGID;
351 }
352
353 /*
354 * Implement correct POSIX saved-id behavior.
355 */
41c20dac
MD
356 p->p_ucred->cr_svuid = p->p_ucred->cr_uid;
357 p->p_ucred->cr_svgid = p->p_ucred->cr_gid;
984263bc
MD
358
359 /*
360 * Store the vp for use in procfs
361 */
362 if (p->p_textvp) /* release old reference */
363 vrele(p->p_textvp);
364 VREF(ndp->ni_vp);
365 p->p_textvp = ndp->ni_vp;
366
367 /*
368 * Notify others that we exec'd, and clear the P_INEXEC flag
369 * as we're now a bona fide freshly-execed process.
370 */
371 KNOTE(&p->p_klist, NOTE_EXEC);
372 p->p_flag &= ~P_INEXEC;
373
374 /*
375 * If tracing the process, trap to debugger so breakpoints
376 * can be set before the program executes.
377 */
378 STOPEVENT(p, S_EXEC, 0);
379
380 if (p->p_flag & P_TRACED)
381 psignal(p, SIGTRAP);
382
383 /* clear "fork but no exec" flag, as we _are_ execing */
384 p->p_acflag &= ~AFORK;
385
386 /* Set values passed into the program in registers. */
387 setregs(p, imgp->entry_addr, (u_long)(uintptr_t)stack_base,
388 imgp->ps_strings);
389
390 /* Free any previous argument cache */
391 if (p->p_args && --p->p_args->ar_ref == 0)
392 FREE(p->p_args, M_PARGS);
393 p->p_args = NULL;
394
395 /* Cache arguments if they fit inside our allowance */
396 i = imgp->endargs - imgp->stringbase;
397 if (ps_arg_cache_limit >= i + sizeof(struct pargs)) {
398 MALLOC(p->p_args, struct pargs *, sizeof(struct pargs) + i,
399 M_PARGS, M_WAITOK);
400 p->p_args->ar_ref = 1;
401 p->p_args->ar_length = i;
402 bcopy(imgp->stringbase, p->p_args->ar_args, i);
403 }
404
405exec_fail_dealloc:
406
407 /*
408 * free various allocated resources
409 */
410 if (imgp->firstpage)
411 exec_unmap_first_page(imgp);
412
413 if (imgp->stringbase != NULL)
414 kmem_free_wakeup(exec_map, (vm_offset_t)imgp->stringbase,
415 ARG_MAX + PAGE_SIZE);
416
417 if (imgp->vp) {
418 NDFREE(ndp, NDF_ONLY_PNBUF);
419 vrele(imgp->vp);
420 }
421
422 if (error == 0)
423 return (0);
424
425exec_fail:
426 /* we're done here, clear P_INEXEC */
427 p->p_flag &= ~P_INEXEC;
428 if (imgp->vmspace_destroyed) {
429 /* sorry, no more process anymore. exit gracefully */
41c20dac 430 exit1(W_EXITCODE(0, SIGABRT));
984263bc
MD
431 /* NOT REACHED */
432 return(0);
433 } else {
434 return(error);
435 }
436}
437
438int
41c20dac 439exec_map_first_page(struct image_params *imgp)
984263bc
MD
440{
441 int s, rv, i;
442 int initial_pagein;
443 vm_page_t ma[VM_INITIAL_PAGEIN];
444 vm_object_t object;
445
446
447 if (imgp->firstpage) {
448 exec_unmap_first_page(imgp);
449 }
450
451 VOP_GETVOBJECT(imgp->vp, &object);
452 s = splvm();
453
454 ma[0] = vm_page_grab(object, 0, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
455
456 if ((ma[0]->valid & VM_PAGE_BITS_ALL) != VM_PAGE_BITS_ALL) {
457 initial_pagein = VM_INITIAL_PAGEIN;
458 if (initial_pagein > object->size)
459 initial_pagein = object->size;
460 for (i = 1; i < initial_pagein; i++) {
461 if ((ma[i] = vm_page_lookup(object, i)) != NULL) {
462 if ((ma[i]->flags & PG_BUSY) || ma[i]->busy)
463 break;
464 if (ma[i]->valid)
465 break;
466 vm_page_busy(ma[i]);
467 } else {
468 ma[i] = vm_page_alloc(object, i, VM_ALLOC_NORMAL);
469 if (ma[i] == NULL)
470 break;
471 }
472 }
473 initial_pagein = i;
474
475 rv = vm_pager_get_pages(object, ma, initial_pagein, 0);
476 ma[0] = vm_page_lookup(object, 0);
477
478 if ((rv != VM_PAGER_OK) || (ma[0] == NULL) || (ma[0]->valid == 0)) {
479 if (ma[0]) {
480 vm_page_protect(ma[0], VM_PROT_NONE);
481 vm_page_free(ma[0]);
482 }
483 splx(s);
484 return EIO;
485 }
486 }
487
488 vm_page_wire(ma[0]);
489 vm_page_wakeup(ma[0]);
490 splx(s);
491
492 pmap_kenter((vm_offset_t) imgp->image_header, VM_PAGE_TO_PHYS(ma[0]));
493 imgp->firstpage = ma[0];
494
495 return 0;
496}
497
498void
499exec_unmap_first_page(imgp)
500 struct image_params *imgp;
501{
502 if (imgp->firstpage) {
503 pmap_kremove((vm_offset_t) imgp->image_header);
504 vm_page_unwire(imgp->firstpage, 1);
505 imgp->firstpage = NULL;
506 }
507}
508
509/*
510 * Destroy old address space, and allocate a new stack
511 * The new stack is only SGROWSIZ large because it is grown
512 * automatically in trap.c.
513 */
514int
515exec_new_vmspace(imgp)
516 struct image_params *imgp;
517{
518 int error;
519 struct vmspace *vmspace = imgp->proc->p_vmspace;
520 vm_offset_t stack_addr = USRSTACK - maxssiz;
521 vm_map_t map = &vmspace->vm_map;
522
523 imgp->vmspace_destroyed = 1;
524
525 /*
526 * Prevent a pending AIO from modifying the new address space.
527 */
528 aio_proc_rundown(imgp->proc);
529
530 /*
531 * Blow away entire process VM, if address space not shared,
532 * otherwise, create a new VM space so that other threads are
533 * not disrupted
534 */
535 if (vmspace->vm_refcnt == 1) {
536 if (vmspace->vm_shm)
537 shmexit(imgp->proc);
538 pmap_remove_pages(vmspace_pmap(vmspace), 0, VM_MAXUSER_ADDRESS);
539 vm_map_remove(map, 0, VM_MAXUSER_ADDRESS);
540 } else {
541 vmspace_exec(imgp->proc);
542 vmspace = imgp->proc->p_vmspace;
543 map = &vmspace->vm_map;
544 }
545
546 /* Allocate a new stack */
547 error = vm_map_stack(&vmspace->vm_map, stack_addr, (vm_size_t)maxssiz,
548 VM_PROT_ALL, VM_PROT_ALL, 0);
549 if (error)
550 return (error);
551
552 /* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the
553 * VM_STACK case, but they are still used to monitor the size of the
554 * process stack so we can check the stack rlimit.
555 */
556 vmspace->vm_ssize = sgrowsiz >> PAGE_SHIFT;
557 vmspace->vm_maxsaddr = (char *)USRSTACK - maxssiz;
558
559 return(0);
560}
561
562/*
563 * Copy out argument and environment strings from the old process
564 * address space into the temporary string buffer.
565 */
566int
567exec_extract_strings(imgp)
568 struct image_params *imgp;
569{
570 char **argv, **envv;
571 char *argp, *envp;
572 int error;
573 size_t length;
574
575 /*
576 * extract arguments first
577 */
578
579 argv = imgp->uap->argv;
580
581 if (argv) {
582 argp = (caddr_t) (intptr_t) fuword(argv);
583 if (argp == (caddr_t) -1)
584 return (EFAULT);
585 if (argp)
586 argv++;
587 if (imgp->argv0)
588 argp = imgp->argv0;
589 if (argp) {
590 do {
591 if (argp == (caddr_t) -1)
592 return (EFAULT);
593 if ((error = copyinstr(argp, imgp->stringp,
594 imgp->stringspace, &length))) {
595 if (error == ENAMETOOLONG)
596 return(E2BIG);
597 return (error);
598 }
599 imgp->stringspace -= length;
600 imgp->stringp += length;
601 imgp->argc++;
602 } while ((argp = (caddr_t) (intptr_t) fuword(argv++)));
603 }
604 }
605
606 imgp->endargs = imgp->stringp;
607
608 /*
609 * extract environment strings
610 */
611
612 envv = imgp->uap->envv;
613
614 if (envv) {
615 while ((envp = (caddr_t) (intptr_t) fuword(envv++))) {
616 if (envp == (caddr_t) -1)
617 return (EFAULT);
618 if ((error = copyinstr(envp, imgp->stringp,
619 imgp->stringspace, &length))) {
620 if (error == ENAMETOOLONG)
621 return(E2BIG);
622 return (error);
623 }
624 imgp->stringspace -= length;
625 imgp->stringp += length;
626 imgp->envc++;
627 }
628 }
629
630 return (0);
631}
632
633/*
634 * Copy strings out to the new process address space, constructing
635 * new arg and env vector tables. Return a pointer to the base
636 * so that it can be used as the initial stack pointer.
637 */
638register_t *
639exec_copyout_strings(imgp)
640 struct image_params *imgp;
641{
642 int argc, envc;
643 char **vectp;
644 char *stringp, *destp;
645 register_t *stack_base;
646 struct ps_strings *arginfo;
647 int szsigcode;
648
649 /*
650 * Calculate string base and vector table pointers.
651 * Also deal with signal trampoline code for this exec type.
652 */
653 arginfo = (struct ps_strings *)PS_STRINGS;
654 szsigcode = *(imgp->proc->p_sysent->sv_szsigcode);
655 destp = (caddr_t)arginfo - szsigcode - SPARE_USRSPACE -
656 roundup((ARG_MAX - imgp->stringspace), sizeof(char *));
657
658 /*
659 * install sigcode
660 */
661 if (szsigcode)
662 copyout(imgp->proc->p_sysent->sv_sigcode,
663 ((caddr_t)arginfo - szsigcode), szsigcode);
664
665 /*
666 * If we have a valid auxargs ptr, prepare some room
667 * on the stack.
668 */
669 if (imgp->auxargs)
670 /*
671 * The '+ 2' is for the null pointers at the end of each of the
672 * arg and env vector sets, and 'AT_COUNT*2' is room for the
673 * ELF Auxargs data.
674 */
675 vectp = (char **)(destp - (imgp->argc + imgp->envc + 2 +
676 AT_COUNT*2) * sizeof(char*));
677 else
678 /*
679 * The '+ 2' is for the null pointers at the end of each of the
680 * arg and env vector sets
681 */
682 vectp = (char **)
683 (destp - (imgp->argc + imgp->envc + 2) * sizeof(char*));
684
685 /*
686 * vectp also becomes our initial stack base
687 */
688 stack_base = (register_t *)vectp;
689
690 stringp = imgp->stringbase;
691 argc = imgp->argc;
692 envc = imgp->envc;
693
694 /*
695 * Copy out strings - arguments and environment.
696 */
697 copyout(stringp, destp, ARG_MAX - imgp->stringspace);
698
699 /*
700 * Fill in "ps_strings" struct for ps, w, etc.
701 */
702 suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp);
703 suword(&arginfo->ps_nargvstr, argc);
704
705 /*
706 * Fill in argument portion of vector table.
707 */
708 for (; argc > 0; --argc) {
709 suword(vectp++, (long)(intptr_t)destp);
710 while (*stringp++ != 0)
711 destp++;
712 destp++;
713 }
714
715 /* a null vector table pointer seperates the argp's from the envp's */
716 suword(vectp++, 0);
717
718 suword(&arginfo->ps_envstr, (long)(intptr_t)vectp);
719 suword(&arginfo->ps_nenvstr, envc);
720
721 /*
722 * Fill in environment portion of vector table.
723 */
724 for (; envc > 0; --envc) {
725 suword(vectp++, (long)(intptr_t)destp);
726 while (*stringp++ != 0)
727 destp++;
728 destp++;
729 }
730
731 /* end of vector table is a null pointer */
732 suword(vectp, 0);
733
734 return (stack_base);
735}
736
737/*
738 * Check permissions of file to execute.
739 * Return 0 for success or error code on failure.
740 */
741int
742exec_check_permissions(imgp)
743 struct image_params *imgp;
744{
745 struct proc *p = imgp->proc;
746 struct vnode *vp = imgp->vp;
747 struct vattr *attr = imgp->attr;
dadab5e9 748 struct thread *td = p->p_thread;
984263bc
MD
749 int error;
750
751 /* Get file attributes */
3b568787 752 error = VOP_GETATTR(vp, attr, td);
984263bc
MD
753 if (error)
754 return (error);
755
756 /*
757 * 1) Check if file execution is disabled for the filesystem that this
758 * file resides on.
759 * 2) Insure that at least one execute bit is on - otherwise root
760 * will always succeed, and we don't want to happen unless the
761 * file really is executable.
762 * 3) Insure that the file is a regular file.
763 */
764 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
765 ((attr->va_mode & 0111) == 0) ||
766 (attr->va_type != VREG)) {
767 return (EACCES);
768 }
769
770 /*
771 * Zero length files can't be exec'd
772 */
773 if (attr->va_size == 0)
774 return (ENOEXEC);
775
776 /*
777 * Check for execute permission to file based on current credentials.
778 */
dadab5e9 779 error = VOP_ACCESS(vp, VEXEC, p->p_ucred, td);
984263bc
MD
780 if (error)
781 return (error);
782
783 /*
784 * Check number of open-for-writes on the file and deny execution
785 * if there are any.
786 */
787 if (vp->v_writecount)
788 return (ETXTBSY);
789
790 /*
791 * Call filesystem specific open routine (which does nothing in the
792 * general case).
793 */
dadab5e9 794 error = VOP_OPEN(vp, FREAD, p->p_ucred, td);
984263bc
MD
795 if (error)
796 return (error);
797
798 return (0);
799}
800
801/*
802 * Exec handler registration
803 */
804int
805exec_register(execsw_arg)
806 const struct execsw *execsw_arg;
807{
808 const struct execsw **es, **xs, **newexecsw;
809 int count = 2; /* New slot and trailing NULL */
810
811 if (execsw)
812 for (es = execsw; *es; es++)
813 count++;
814 newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK);
815 if (newexecsw == NULL)
816 return ENOMEM;
817 xs = newexecsw;
818 if (execsw)
819 for (es = execsw; *es; es++)
820 *xs++ = *es;
821 *xs++ = execsw_arg;
822 *xs = NULL;
823 if (execsw)
824 free(execsw, M_TEMP);
825 execsw = newexecsw;
826 return 0;
827}
828
829int
830exec_unregister(execsw_arg)
831 const struct execsw *execsw_arg;
832{
833 const struct execsw **es, **xs, **newexecsw;
834 int count = 1;
835
836 if (execsw == NULL)
837 panic("unregister with no handlers left?\n");
838
839 for (es = execsw; *es; es++) {
840 if (*es == execsw_arg)
841 break;
842 }
843 if (*es == NULL)
844 return ENOENT;
845 for (es = execsw; *es; es++)
846 if (*es != execsw_arg)
847 count++;
848 newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK);
849 if (newexecsw == NULL)
850 return ENOMEM;
851 xs = newexecsw;
852 for (es = execsw; *es; es++)
853 if (*es != execsw_arg)
854 *xs++ = *es;
855 *xs = NULL;
856 if (execsw)
857 free(execsw, M_TEMP);
858 execsw = newexecsw;
859 return 0;
860}